Machine Learning Functions
Machine learning functions let you work with your data set in different stages of the data analysis process:
- Preparing models
- Training models
- Evaluating models
- Applying models
- Managing models
Some Vertica machine learning functions are implemented as Vertica UDx functions, while others are implemented as meta-functions:
- A UDx function accepts an input relation name from a
SELECTstatement that calls the functions is composable—it can be used as a sub-query in another
- A meta-function accepts the input relation name as a single-quoted string passed to it as an argument or a named parameter. The data that the
SELECTstatement returns cannot be used in a sub-query. Machine learning meta-functions do not support temporary tables.
All machine learning functions automatically cast NUMERIC arguments to FLOAT.
Before using a machine learning function, be aware that any open transaction on the current session might be committed.